On this page
Short answer
We use optical imagery for spectral and visible context when usable clear observations exist, and consider SAR where cloud or illumination limits that record. Radar adds different evidence, not a universally better picture. Sensor geometry, season, processing and independent observations must be checked before classifying a change.
- Scheduled revisit is not the same as a usable cloud-free observation.
- SAR backscatter can change with moisture, roughness and geometry, even without the change you are looking for.
- A candidate change needs a dated reference and an explicit validation test.
Who this is for: Infrastructure, environmental and GIS teams planning a satellite change-monitoring workflow.
Different sensors answer different questions
Optical multispectral imagery records reflected energy across spectral bands. It can provide useful land-cover and visible context, but cloud and shadow can remove the observations that matter for an event. A composite assembled from several dates is not necessarily a single-date view.
SAR supplies its own radar illumination. It can acquire observations through cloud and without sunlight, but its side-looking geometry and scattering response need specialist interpretation. Moisture, surface roughness, polarization and viewing angle influence the signal. We avoid treating radar images as ordinary photographs.
Technical references: Copernicus Data Space: Sentinel-2 documentation; NASA/JPL: what is imaging radar?; NASA: Get to Know SAR.
Choose the evidence around the decision
This table is our planning synthesis. The actual archive must be inspected before promising a monitoring interval. A nominal revisit interval does not establish that the area has a valid, comparable observation on every visit.
| Decision | Optical contribution | SAR contribution | Check first |
|---|---|---|---|
| Land-cover or vegetation change | Spectral and visual context | Complementary structural or moisture response | Comparable seasonal stage and dates |
| Water extent during a cloudy period | Clear observations where available | Cloud-tolerant acquisition | Permanent water, shadow and urban or vegetated areas |
| Infrastructure surroundings | Visible context where objects can be resolved | Complementary surface/scattering changes | Object size, image alignment and independent evidence |
Technical references: Copernicus Data Space: Sentinel-2 documentation; NASA: Get to Know SAR; NASA: HydroSAR Sentinel-1 classifications.
Three ways a change map can mislead
For flood interpretation, vegetation, developed areas and steep terrain complicate SAR classification. The linked NASA HydroSAR material is framed as situational awareness, not a definitive property-level impact assessment. We keep that distinction when designing a review workflow.
- A seasonal comparison: vegetation, soil moisture or farming stage differs between dates. The map may be detecting that difference rather than the intended event.
- A geometry problem: misregistration, radar shadow, foreshortening or layover affects the comparison, especially around slopes and structures.
- A label problem: a dark radar response or changed pixel is automatically called water or damage without checking the setting and independent evidence.
Technical references: NASA: Get to Know SAR; NASA: HydroSAR Sentinel-1 classifications.
A reviewable change-monitoring workflow
- Define the change, area, minimum useful object size and decision to be supported.
- Inspect the actual archive for dates, coverage, processing levels and unusable observations.
- Choose comparable inputs and document cloud, shadow or other exclusion masks.
- Check spatial alignment and keep a consistent comparison method.
- Review candidate changes with the observation context visible beside them.
- Validate a sample against dated independent evidence, and report errors and unobserved areas.
Technical references: Copernicus Data Space: Sentinel-2 documentation; NASA: Get to Know SAR.
What should a monitoring report show?
We would include acquisition dates, processing levels, pixel size, masks, alignment checks and the change definition. Mark areas with no valid observation separately from areas where no change was detected. State whether findings were checked against field records or remain candidates for investigation.
A map is more useful when a reviewer can trace a finding to its source observation. Confidence scores, where produced by a model, should be interpreted alongside an independently assessed validation sample.
Technical references: NASA: HydroSAR Sentinel-1 classifications.
Common questions
Can SAR see through every obstruction?
No. SAR can acquire through cloud without sunlight, but shadow, geometry, surface properties and the radar wavelength affect what can be observed. Do not interpret cloud tolerance as a guarantee about every feature.
Does a cloud-free mosaic show one event date?
Not necessarily. A multi-date mosaic can combine observations from different times. Check the time information before using it to describe a specific event.
Is a changed satellite pixel evidence of infrastructure damage?
Not by itself. Check the change definition, season, sensor geometry, alignment and independently dated evidence before making a damage claim.
Official references
Sources checked 1 November 2026. The linked publishers remain the authority for their specifications and services.
- NASA: Get to Know SAR
Radar scattering, polarization and side-looking geometry.
- NASA/JPL: what is imaging radar?
Historical explanation used for stable radar physics, not current mission-status claims.
- Copernicus Data Space: Sentinel-2 documentation
Optical products, valid observations and cloud/shadow treatment in mosaics.
- NASA: HydroSAR Sentinel-1 classifications
Interpretation limits and supporting evidence for SAR flood classifications.